0

I've recently encountered the claim that a confidence interval is 1-$\alpha$, (where I assume $\alpha$ is a p-value).

e.g. that a p-value of 0.05 is equivalent to a 95% confidence interval.

I believe this claim to be false, as I've recently read that p-values are not predictive, but I wonder if a statistician (which I am not) could shed some light on this?

Thanks

Edd Barrett
  • 153
  • 1
  • 7
  • 1
    What do you mean by "p-values are not predictive"? – Dave Sep 12 '19 at 10:33
  • 2
    $\alpha$ is not the same as p-value. https://www.thoughtco.com/the-difference-between-alpha-and-p-values-3126420 – user2974951 Sep 12 '19 at 10:41
  • @Dave That you can't interpret a p-value as a probability indicator. I read this in the book "the skeptics guide to the universe". The author says "A p-value of 0.05 indicates a 5-percent probability that the data are due to chance than a real effect. Except that's not actually true" and then later "A study with a p-value of 0.01 may only have a 50 percent chance not the 99 percent that most people would assume. Put another way, people (even experienced scientists) tend to think of the p-value as a predictive value, but it isn't". I'm confused because user2974951's link states the opposite. – Edd Barrett Sep 12 '19 at 11:07
  • I think you are right that I confused the p-value with the $\alpha$ though. – Edd Barrett Sep 12 '19 at 11:09
  • @EddBarrett The p-value is something you calculate based on the data, perhaps by conducting a t-test of the mean. The $\alpha$-level is a value that you pick. If $p$ falls below this $\alpha$ threshold, then you have a sufficiently unlikely test statistic to believe your null hypothesis, and reject the null in favor of your alternative. The reason is that, assuming the null hypothesis is true, the calculated value is just too extreme to believe that it comes from the circumstances of the null hypothesis, like 999 flips of heads and 1 tails---probably NOT a fair coin! – Dave Sep 12 '19 at 11:42
  • @Dave I see, makes sense. In light of this, my question is kind of void. What should I do? Would you like to write an answer? – Edd Barrett Sep 12 '19 at 12:34
  • @EddBarrett This is a duplicate of another question. Here is a link to the accepted answer: https://stats.stackexchange.com/a/169149/247274. – Dave Sep 12 '19 at 17:53

0 Answers0